Financial network communities and methodological insights: a case study for Borsa Istanbul Sustainability Index

Understanding market dynamics necessitates a deep understanding of financial networks, which offer a detailed view of relationships and interactions between market entities. Analyzing these networks, constructed using measures like price correlations and trading volumes, reveals market structure and the interplay of various factors. This allows for the identification of clusters of interconnected stocks (communities) often corresponding to specific sectors, highlighting sectoral interdependencies and information flow. Financial networks also reveal key nodes or hubs, crucial for understanding systemic risk. Recently, Environmental, Social, and Governance (ESG) factors have become vital in financial decision-making, encompassing criteria assessing a company's performance beyond traditional financial metrics. The rising importance of ESG reflects a shift toward sustainable investing, with studies showing that companies with strong ESG performance manage risks better and achieve long-term success. Despite extensive research on financial networks, understanding how ESG factors influence network formation and dynamics remains a significant gap. This study aims to bridge this gap by investigating the impact of ESG scores on clustering and community formation within various financial network models, integrating sustainability metrics to provide a more comprehensive understanding of market dynamics and uncover how environmental, social, and governance considerations influence intercompany relationships and community formation.

The literature review synthesizes insights from recent studies on ESG factors and financial networks. Several studies demonstrate a positive correlation between ESG scores and improved financial performance (Kim and Li, 2021; Sinha et al., 2019; Aybars et al., 2019; Caporale et al., 2022; Luo et al., 2024). Batrancea et al. (2022c, 2022a, 2022b, 2021, 2023b, 2023a) highlight the importance of governance quality and social policies in fostering economic growth, with varying degrees of focus on environmental sustainability. Research on ESG integration in investment strategies indicates that incorporating ESG signals can lead to comparable historical performance to traditional factor portfolios (Chan et al., 2020; Zaccone and Pedrini, 2020; Ma, 2023; Park and Lee, 2023; Park and Oh, 2022). Studies also link higher ESG rankings to enhanced financial stability by mitigating individual and systemic risks (Stolbov and Shchepeleva, 2022; Lupu et al., 2022; Ling et al., 2023). The relationship between ESG and market value is also explored, with some studies showing positive relationships (Ionescu et al., 2019; Lee and Isa, 2023; Yoon et al., 2024), while others find no significant direct impact (Junius et al., 2020). Research on ESG in the context of Turkey focuses on the relationship between monetary policy and financial asset returns (Kartal et al., 2024b), the link between ESG disclosures and scores (Kartal et al., 2024a), and the influence of climate risk (Pata et al., 2024). Studies on ESG's predictive power show that integrating ESG metrics into predictive models can improve forecasting accuracy (Ang et al., 2023; Giese et al., 2019; Lee et al., 2024; Gong et al., 2024). Methodological approaches in ESG research involve both quantitative and qualitative metrics, with increasing use of machine learning techniques (Meiden and Silaban, 2023; Escrig-Olmedo et al., 2017; Efimova et al., 2021; Raza et al., 2022; Zhang and Zhang, 2023).

This study employs five distinct network construction approaches: correlation networks, continuous mutual information networks, discrete mutual information networks, linear causality networks, and nonlinear causality networks. Each approach provides a unique perspective on the relationships between companies within the Borsa Istanbul Sustainability Index (XUSRD). **Correlation Networks:** These networks represent the linear relationships between daily closing prices of companies. Logarithmic returns are calculated using the formula: Cl(t) = log P₁(t) - log P(t-1). Pearson correlation coefficients (pᵢⱼ) are computed, and the correlation distance is defined as dc(i,j) = √2(1-pᵢⱼ). The triangulated maximally filtered graph (TMFG) approach is used to filter out insignificant connections. **Mutual Information Networks:** Mutual information quantifies the amount of shared information between two time series. Both continuous and discrete bivariate mutual information are calculated. Continuous mutual information (I_c) is computed using kernel density estimation techniques, while discrete mutual information (I_d) is calculated after discretizing the logarithmic returns. The weight functions used are 1/I_c and 1/I_d, respectively, with TMFG filtering applied. **Causality Networks:** Linear causality is assessed using Granger causality tests, examining whether past values of one company's returns significantly predict another's. Nonlinear causality is investigated using Transfer Entropy, which measures the information flow from one time series to another, considering nonlinear dependencies. Weights are assigned to edges based on p-values, with lower p-values indicating stronger causal relationships. TMFG filtering is applied to both linear and nonlinear causality networks. **Community Detection:** Two community detection algorithms are employed: the Leading Eigenvector method, based on the eigenvectors of the modularity matrix, and the Girvan-Newman algorithm, based on edge betweenness. Both methods are applied to each of the five network types. Modularity is used as an evaluation metric for community quality. **Network Comparison:** Five standard community comparison metrics—Normalized Mutual Information (NMI), Variation of Information (VOI), Split Join Distance (SJD), Rand Index (RI), and Adjusted Rand Index (ARI)—are used to assess the similarities and differences between communities identified by different methods. A novel comparison methodology based on random walks is also introduced to further analyze network interdependencies.

The study analyzed daily closing prices of 78 companies listed on the Borsa Istanbul Sustainability Index (XUSRD) from 2023. A linear regression model (Adjusted R-squared = 0.9402) indicated that Resource Use, Innovation, Human Rights, Product Responsibility, Workforce, Community, Management, Shareholders, and CSR Strategy positively and significantly influence ESG scores. ANOVA results further confirmed the significance of these factors. Network analysis revealed distinct clustering patterns across different network models. Correlation networks highlighted strong sectoral linkages, particularly within the finance and manufacturing sectors. Mutual information networks (both continuous and discrete) revealed both intra- and inter-sectoral dependencies, with continuous mutual information particularly revealing strong linkages between financial and manufacturing sectors. Causality networks (linear and nonlinear) revealed predictive relationships in stock price movements, with both linear and nonlinear causality showing significant clustering within the financial and manufacturing sectors. However, nonlinear causality revealed more intricate cross-sectoral interdependencies. Community detection using the Leading Eigenvector and Girvan-Newman algorithms revealed distinct community structures within each network type. Both algorithms consistently identified strong sectoral clustering, particularly within the financial and manufacturing sectors. However, the specific composition of communities varied slightly between the two algorithms, indicating different perspectives on the underlying relationships. The novel random walk-based comparison method revealed significant transfer entropy between different network types. For example, the nonlinear causality network received high transfer entropy from the linear causality network, suggesting that linear causality captures important elements of the dynamics that are further elaborated in nonlinear interactions. Nonparametric tests (Kruskal-Wallis, Conover, and log-rank) were used to assess the impact of ESG scores on community formation. These tests revealed significant influences of specific ESG factors on community structures across different network models. For instance, Resource Use, Innovation, Human Rights, Product Responsibility, and CSR Strategy showed significant associations with community membership in various network types. The results suggest companies with strong performance in these areas tend to form more cohesive communities, highlighting the role of sustainability in shaping financial networks.

The findings address the research question by demonstrating the significant influence of ESG scores on community formation within financial networks. The results show that sustainability metrics are not merely individual company performance indicators but also shape broader market dynamics by fostering interconnected communities of companies with similar ESG practices. The significance of various ESG factors across different network models highlights the multifaceted nature of sustainability's influence on financial markets. For example, the significant impact of Resource Use in correlation networks suggests that companies with similar resource management practices are more likely to experience correlated stock price movements. The significant influence of Innovation and CSR Strategy in other network types suggests that these factors drive both intra-sectoral and cross-sectoral interdependencies. These findings have significant implications for sustainable investment practices, suggesting that investors can use ESG factors to identify companies that are likely to be more resilient and less risky. The integration of ESG factors into financial network analysis enhances our understanding of market dynamics and promotes sustainable investment strategies. The results highlight the need for investors and policymakers to consider ESG factors when making investment decisions to foster more resilient and socially responsible financial markets.

This study provides valuable insights into the role of ESG factors in shaping financial networks. It demonstrates the significant influence of sustainability metrics on community formation, highlighting the interconnectedness between ESG performance and market dynamics. The findings underscore the importance of integrating ESG considerations into financial analysis and investment decision-making to promote sustainable and responsible financial markets. Future research could explore the dynamic evolution of these networks over time, incorporate alternative ESG data sources, and investigate the interactions between ESG factors and other financial metrics.

This study has several limitations. First, it relies solely on daily closing prices, potentially overlooking intraday variations. Second, the ESG scores used were based on publicly available data, which may not fully reflect companies' sustainability performance. Third, the static nature of the network models does not capture temporal changes in company relationships. Fourth, the study focused on a specific index, limiting generalizability to other markets. Finally, the nonparametric tests, while robust, may not fully capture the complex interactions between ESG factors and network communities.